A Modelling Approach to Support the Management of Flood and Pollution Risks for Extreme Events in Urban Stormwater Drainage Systems Bryan Ellis*, Christophe Viavattene+, Michael Revitt*, Christian Peters# and Heiko Seiker# *Urban Pollution Research Centre, Middlesex University, London, UK. + Flood Hazard Research Centre, Middlesex University, London. UK # IPSmbH, Hoppegarten, Berlin, Germany. Corresponding Author:
[email protected]
Abstract The complexity and interactions of flooding and associated pollution resulting from exceedance surface water flows during extreme storm events presents a major threat for future sustainable urban drainage management. Innovative coupled 1D/2D modelling approaches are described for the detailed delineation and analysis of surface flowpaths, flood depths and velocities during such extreme events. The modelling structure and potential outputs are illustrated by reference to the Birmingham Eastside SWITCH demonstration area.
Keywords: urban stormwater, drainage complexity, exceedance flows, GIS-based 1D/2D modelling
1
Introduction
The principal objective of Work Package (WP) 2.1 “Technological Options for Stormwater Control Under Conditions of Uncertainty” within the SWITCH project is to utilise the control and treatment of impermeable surface runoff to support sustainable urban stormwater management and provide an enhanced level of service and quality of life to the urban community. Both current and future threats and uncertainties to such holistic, sustainable stormwater management were identified in consultation with the Birmingham and Belo Horizonte Learning Alliances (LAs) and a guideline risk assessment strategy developed and successfully tested with stakeholders. This work has been fully described in Deliverable 2.1.1b (Ellis et al., 2008a), with the methodological application further illustrated in a paper presented at the 2nd SWITCH Scientific Meeting in Tel Aviv, Israel (Scholes et al., 2007). The primary threat for future urban stormwater control identified by both SWITCH demonstration cities was that of increasing flood risk occurring from inadequate drainage capacity (of both carrier sewers and receiving water channels), increasingly exacerbated by urban creep and climate change. A risk assessment matrix for flooding and associated pollution risks, developed in conjunction with the LAs for both contemporary and future conditions, was outlined in a paper presented at the 3rd SWITCH Scientific Meeting convened
in Belo Horizonte, Brazil (Ellis et al., 2008b). One major issue and source of uncertainty highlighted in this risk analysis work is that of accurately defining and targeting appropriate mitigating measures for the relief of extreme event flood and pollution “hotspots” within urban development areas. This challenge comprises the central focus of Task 2.1.3 within SWITCH WP 2.1 in terms of quantifying the potential of different stormwater control options through the application of a GIS-based modelling approach. GIS systems are most commonly used to collate and manage spatial data required as essential input for stormwater models such as SWMM, MIKE Urban CS, InfoWorks SD or STORM whose comparative attributes and strengths have been reviewed in Deliverable 2.1.2, Part A (Scholes and Revitt, 2008). In the context of a typical urban development scenario involving multiple stakeholders having a wide variety of interests and concerns, there is clear potential for the use of a central data integration and communication tool to act as a precursor to analytical modelling. In addition, the availability of such a forecasting/screening tool is regarded as being a priority warning system for future pluvial flood management in the UK (Falconer et al., 2009) given the change in national perspectives on flood risk management following the extreme storm events of 2007. The development of this type of specific GIS tool, which facilitates stakeholder involvement and interaction in the decision-making process of BMP selection and location, is an identifed objective of SWITCH WP 2.3 (Task 2.1) and the development of this GIS tool has been reported elsewhere (Viavattene et al., 2008). The application of such a GIS-based tool in combination with a stormwater modelling tool to provide a novel, coupled 1D/2D approach to the analysis and quantification of complex interurban extreme event flooding and pollution represents a further novel contribution within SWITCH and has been briefly introduced in Ellis et al., (2008b). The development of an extreme event management approach to map potentially vulnerable urban areas in order to prioritise mitigation measures and maximize lead times for emergency flood preparation, as well as underpinning Stormwater Management Plans (SWMPs), is regarded as constituting an urgent priority need in the UK (Pitt, 2008). Such risk mapping of potentially vulnerable urban areas is also a prime requirement of the EU Floods Directive. In addition, it would also provide a service tool of value to drainage practitioners and local authorities responsible for urban surface water management. This current paper develops this theme further and outlines the challenges posed by drainage system complexity and describes both the structure of the 1D/2D modelling approach and how it might be applied to sites and sub-catchments of the SWITCH demonstration cities.
2
Complexity and interactions in urban surface drainage
It is now generally acknowledged that urban flooding and pollution are frequently the result of multiple urban land use sources associated with a combination of overland flow, sewer surcharging and receiving watercourse overloading. Figure 1 provides detail on the sources, pathways, receptors and return flows for the urban drainage system and emphasises the complexity of interactions between the various above and below ground sources. The intraurban response to flooding operates through various process mechanisms and acts on differing spatial scales, combining above and below ground systems, storage facilities and flow routes. Four individual but interlinked drainage systems can be identified from Figure 1: -
a foul (combined) sewerage system with combined sewer overflows (CSOs) discharging to the receiving water,
Rainfall Overland flow during wet weather
Impermeable Surfaces
Permeable Surfaces Sheetflow
Private Drains
Groundwater
Highway Drains
Combined Sewers
Surface Water Sewers Exceedance flows
SWOs
Infiltration
Ditches and Culverts
Springs and Hyperheic Flows
Inflow SWOs CSOs CSOs
Sewerage and Water Treatment
Sub-Catchment Receiving Waters
Regional Catchment Receiving Water
Figure 1. Interactions Between Urban Drainage Sources, Pathways and Receptors. -
a separate surface water sewer system with surface water outfalls (SWOs) discharging to the receiving water, the receiving water (normally “heavily modified”), and exceedance surface flows during extreme wet weather conditions.
Interconnections, including system cross- (or mis-) connections, infiltration and inflow pathways as well as system abstractions further complicate the process interactions. A theoretical distinction can be made between pluvial flooding caused by rainfall-runoff over impermeable surfaces and exceedance flooding caused by a combination of surcharging, overland flow and sheetflow from sewers, impermeable and saturated permeable surfaces, as well as over-top flooding from ditches and culverts during extreme wet weather conditions. These distinct source contributions are identified in Figure 1 although it is practically impossible to separately identify them under actual field conditions. During extremely intense rain storms, pluvial flooding may occur on the urban surface even when the sewer network is only subject to free surface (non-surcharging) conditions i.e under non-exceedance conditions as the roadside gullies cannot pass the surface runoff fast enough into the belowground sewer system. However, the interactive nature of urban drainage systems demands a fully integrated, GIS-based modelling approach to simulate a replication of the real flooding and pollution situation during extreme events (>1:30 RI). Potential responses to the flood and associated pollution driver mechanisms must take into consideration this complexity of sources and scales of operation. Control and management approaches should therefore consider the level of the individual building (and curtilage), through the plot, site and subcatchment levels with initial data input to the modelling process giving indications of flood mechanisms and interactions, the areal distribution and frequency of flooding as well as
damage consequences. A crucial data component is that relating to road gully and manhole location, spacing and surcharging contributions to surface overland flow during storm events. Figure 2 illustrates the complex hydraulic interactions between major (overland) and minor (sewer) drainage systems that can occur under exceedance flow conditions during a storm event and which can lead to “coincident” flooding. The varying storm design standards shown for the differing parts of the sewer system further illustrate how the hydraulic capacity of the minor system is readily overcome during extreme events with roadside gully chambers, normally designed to a 1:1 - 1:2 RI capacity, being rapidly drowned out and contributing to exceedance flows in the highway cross-section. Extreme event impermeable/permeable surface runoff 1:1/1:2
A proportion of flow enters the minor pipe system
Continued exceedance flow conveyed over the surface
Combination of flooding from minor system and overland flow in the major system results in surface and property flooding Surcharging of minor system when system capacity exceeded or outfall blocked
1:25/1:30
Minor sewer system flow During extreme events Major system: above ground Minor system; below ground
1:75/1:100
Sewer continuation to downstream system and outfall
Rise in river level
Figure 2. Urban Drainage System Extreme Event Interactions (Based on Digman et al., 2006) The key control interaction for management action is the ability to discharge excess flows generated within the development site to appropriate temporary storage and/or infiltration facilities in order to reduce surface flooding and to improve the final discharge quality. In addition, above-ground flood routes and temporary storage for extreme event exceedance flows need to be delineated rather than seeking to expand or enlarge traditional below-ground conveyance systems given the prohibitive costs of system rehabilitation and enlargement. The complexity of urban drainage systems can be further illustrated by reference to the July 2002 floods in Glasgow, Scotland where post-flood modelling studies attributed the discrete catchment flood volumes to the four specific sources mentioned above (Adshead, 2007). Pluvial overland runoff from impervious surfaces was considered to contribute over one-third (34%) of the total flood volume with sewer surcharging accounting for 23% and watercourse (fluvial) overspill some 43%. The preliminary Pitt review of the summer 2007 UK floods attributed some two thirds of the flooding to inadequacies in the current hydraulic capacity of surface water drainage systems (Pitt, 2007) and the final report (Pitt, 2008) drew attention to the need for appropriate modelling tools. Considerable effort is now being placed on developing integrated, interactive surface and sub-surface flood models for exceedance conditions in urban areas and a number of commercial modelling approaches such as TELEMAC (www.telemacsystem.com), TUFLOW (www.tuflow.com), FLO-2D (www.flo2d) and FloodArea (www.savannah-simulations.com) have all recently appeared which attempt to conceptualise surface drainage complexity and interactions.
3
Integrated modelling approaches for extreme events
A number of recent reviews within the UK have now reported on the challenges presented by complex flooding and diffuse pollution sources and mechanisms (Balmforth et al., 2008; Dignam et al., 2006; Pitt, 2008). The UK government Foresight report (Evans et al., 2004) estimated that the number of urban properties at risk of pluvial flooding could be up to 0.5 million by the 2080s. It is estimated that two-thirds of the 57,000 flooded homes during the recent summer 2007 extreme events were inundated by surface water drainage with damage estimated at about £3 billion (Crichton, 2008), In principle these findings provide a considerable stimulus to facilitate the availability of robust and accurate modelling techniques for assessing the real-time risk of such complex exceedance flow conditions. The availability and implementation of fully integrated coupled models for urban drainage will enable decision-makers to identify the likely severity of flood and pollution impacts as well as to select and locate compensatory storage and/or infiltration controls (Diaz-Niets et al., 2008). High-level screening for exceedance urban flooding would also serve to inform preliminary flood risk assessment (FRA) as required under the EU Floods Directive. The majority of modelling approaches for surface water flooding have been based on 1D overland routing of an assumed uniformly distributed design rainfall event (or “blanket” approach) although there is increasing interest in dynamic time series rainfall. A decoupled 1D sewer model with 2D overland routing such as SIPSON (Djordjevic et al., 2005), produces a much better estimate of the spatial extent of flooding as 2D solutions are not forced to follow pre-defined flowpaths. However, 1D/2D modelling of overland surface flows is still in its infancy. Such 1D/2D approaches utilizing GIS and high-resolution digital terrain (or elevation) modelling (DTM or DEM) obtained with LiDAR or by photogrammetry, can handle large data sets and provide a very powerful tool for detailed urban surface water management. DTM/DEM dictates the scale to which the modelling can be undertaken and the accuracy of the output; for detailed extreme flow (Type III) risk assessment, fine scale (±50 – 150 mm) topographic resolution is necessary. Only such 2D modelling can deliver accurate strategic planning information on the detailed temporal distribution of flood depth, duration and velocities during an extreme event. An example of 1D/2D modelling output for identifying surface flow paths and depths during an extreme event was illustrated in the paper presented at the 3rd SWITCH Scientific Meeting (Ellis et al., 2008b). The surface exceedance flows and flowpaths are mapped by a 2D ”rolling-ball” routing algorithm which tracks overland flow paths down the steepest gradient from flooded gullies and manholes along preferred highway “channels”, and replicates the actual physical process of surface flooding (Hankin et al., 2008), based on ground topography of the road and containing kerbs, walls etc. Where overland flows are constrained by such kerb lines, walls etc., it may be entirely adequate to model the flow paths in 1D. However, such a 1D approach confines the overland flow to linear “channels” defined by the street lines and does not account for varying water depths or the filling of low points (sinks or ponds) which can lead to changes in flow route direction. The 1D approach has been now largely superseded by 2D overland flow modelling which allows for multiple and multi-directional flow paths with uncertainties reduced by the small cell sizes which can be used in the modelling analysis. The topography, slope and surface characteristics (such as roughness) are used to calculate how water will flow and spread across the surface. This approach can overestimate the actual flood ponding as it actually occurs on the ground as it does not allow for surface flows to return back into the below-ground system as may often occur, particularly as the storm event subsides. However, a coupled sewer modelling approach provides for a more realistic interchange of flows between the various types of drainage system as well as being able to take into account receiving water (fluvial) flood effects. The modelled output can then allow a determination of the routes of least impact (or damage), so that overland flows can be safely diverted for attenuation, infiltration and/or treatment. It is important to
note however, that the individual contributing flow sources to the final flood outcomes are not distinguished in the modelling simulations. Such integrated modelling, being based on the inter-connectivity of the different sources of urban drainage and their effect on intra-urban flooding and pollution, provides a much firmer strategic foundation for risk management. The dynamic surface and sub-surface hydraulic modelling techniques allow the flow complexity to be spatially and temporally analysed and provides a much better understanding of the extreme event problem and the management of potential solutions including the location of sacrificial flood storage areas (Bamford et al. 2008) based on source contributions. However, such modelling requires detailed data on the geo-located flooding and the generating rainfall event, which is not always available or requires considerable effort and cost as well as rigorous “ground truthing” (Gill, 2008). One particular problem relates to the accuracy of the digital elevation model (DEM) used in the modelling algorithms as walls, fences, alleyways, driveways (and sometimes bridges, flyovers etc.) are not always represented in the DEM. Unless they are filtered from, or taken into account in the data set, such common urban features can lead to apparent shortcutting and/or blocking of surface flow routes within the model results (Boonya-aroonnet et al., 2007). However, recent work on automated “barrier” detection to surface flow path analysis has suggested that utilising LiDAR data representation based on recognition of elevated map features (such as bridges, crossings etc.), can provide the possibility of much more effective flowpath mapping (Evans, 2008). Nevertheless, fully dynamic risk mapping is only likely to be cost-effective for flood “hotspots” of high vulnerability and/or damage costs and will require careful “ground truthing” (Falconer et al., 2009). The hydraulic modelling also requires a supporting tool to identify appropriate locations for, and types of, BMP/SUDS that might be retrofitted and dimensioned for flow and quality control. Such a tool is being developed concurrently within the SWITCH programme and its structure and performance capabilities have been outlined elsewhere (Viavattene et al., 2008; Ellis et al., 2008a). There are additional pressures on the spatial and process complexity of exceedance flooding associated with climate change, future urban development and urban creep. Climate change is predicted to increase winter rainfall by 10 – 30% in the UK by the 2080s with rainfall intensities increasing by up to 20%. Up to 3 million new dwellings are scheduled to be built in England alone by 2016 and “urban creep” from paving of previously pervious areas will collectively increase significantly the surface water flood risk. It is therefore appropriate in the context of SWITCH demonstration cities to consider user-focussed support tools to identify the threats and hazards posed by future urban exceedance flows and to provide a basis for the optimum choice and location of mitigating and emergency measures.
4
The structure of 1D/2D modelling approaches
Recent linkages between modelling tools and platforms have significantly improved and now allow user-specific linkages to be created between very different and independent modelling approaches (a process called Open Modelling Interface or OpenMI). 2D approaches which resolve the continuity and momentum equations for shallow water free surface flow are now readily available and which can additionally incorporate the full functionality of 1D hydrodynamic network analysis (such as STORM, SWMM etc.) for sewered and surface channel flows. Figure 3 shows the nodes and links of the Birmingham Eastside development area surface water sewers as modelled by STORM and which
Figure 3. STORM modelling ouput for Birmingham Eastside provides a basis for the 1D hydrodynamic input to such a coupled approach. 1D domains include the surface water sewer network and nodes, open-channel ditches and culverts with overland 2D domains extending over the surface from gully and manhole surcharge locations. Stormwater sewer data and urban land use types areprovided via a GIS layer and 2D domains derived from DEM/DTM (±0.05 m) in unfiltered or filtered form i.e buildings, vegetation, bridges etc, being removed from the data. The OS Mastermap street network is ”stamped” onto unfiltered DEM/DTM, thus forcing the road network to act as a part of the surface drainage system. In this way, even very minor features such as 100 mm kerbs become significant controlling features affecting flood pathways and storage points. Each land use type (or material) is assigned a roughness coefficient which can vary to allow for frictional losses due to bridge piers, walls, surface composition etc., which can be obtained from conventional “look-up” tables for Manning “n” values. The clearest identification of flood risk domains (or impact zones; IZs) within a development site can be obtained through applying a contour polyline screening (CPS) technique based on the raster DEM/DTM grid mesh. Each pixel in the IZ grid is treated as an individual storage or impact cell (IC) which is assigned an elevation from the raster DEM and closed contours are constructed using automatic ArcMap functions (Figure 4). CPS uses filtered topographic data to create polygons representing closed contours within the IZ using standard GIS functions. These unfiltered closed contours form a sequence of “rings” or depth contours around sinks (depressions or ponds) as seen in Figure 4. The identified IZ depressions (or ponds) can then be filtered with minor shallow-depth hollows (
